Show HN for July 31, 2025

Hey Pietro and Luigi here, we are the authors of mcp-use (https://github.com/mcp-use/mcp-use).

When the first MCP servers came out we were very excited about the technology, but as soon as we wanted to get our hands dirty, we found out that MCP could be used only through Claude Desktop or Cursor. As engineers, we did not like that. MCP seemed like something you wanted to use to build products and applications yourself, not something to hide behind a closed source application.

So we approached the SDK but were pretty dissatisfied with the developer experience (double async loops, lots of boilerplate). We decided to write mcp-use to make our lives easier.

mcp-use lets you connect any LLM to any MCP server in just 6 lines of code. We provide a high level abstraction over the official MCP SDK that makes your life easier and supports all the functionalities of the protocol.

Demo video here: https://www.youtube.com/watch?v=nL_B6LZAsp4.

The key abstractions we provide are called MCPClient and MCPAgent.

MCPClient takes in a set of server configurations, automatically detects the transport type and creates a background task which handles the stream from/to the server.

MCPAgent is a combination of the MCPClient, an LLM, and a custom system prompt. It consumes the MCP client by transforming the tools, resources and prompts into model agnostic tools that can be called by the LLM.

The library also contains some cool utilities:

- secure sandboxed execution of MCP servers (we know the protocol doesn't shine for security)

- meta-tools that allow the agent to search over available servers and tools (to avoid context flooding) and connect dynamically to the server it needs (you could create the omnipotent agent with this).

Some cool things we did with this: - write an agent that can use a browser and create/read linear tickets updated with latest information on the internet

- write an agent that has access to the metrics of our company to automatically create weekly reports.

- I connected an agent to an IKEA curtain I hacked an MCP on to adapt the lighting of my room from images of the lighting situation.

- recreated am open source claude code like CLI, with full MCP capability but with custom models and BYOK (https://github.com/mcp-use/mcp-use-cli).

We recently crossed 100,000 download and we are used by many organizations, including NASA!

We’d love to hear what you think of it, most importantly how we can improve it! We are happy to answer any questions and look forward to your comments.

Hey HN, we're Haakam, Michael, and Adi. We're building AgentMail (https://agentmail.to/), an API to give AI agents their own email inboxes. We’re not talking about AI for your email, this is email for your AI.

We started building email agents because they can converse with users in their inboxes, automate email-based workflows, and authenticate with third-party applications. Given these unique capabilities, we think email will be a core interface for agents.

But we were building on top of Gmail, which was a struggle: poor API support, expensive subscriptions, rate limits, sending limits, GCP Pub/Sub, OAuth, crappy keyword search, and an overall terrible developer experience.

Gmail and other providers didn’t work for us. So we decided to bite the bullet and build our own.

AgentMail is like Gmail, but API-first, with programmatic inbox creation, events over webhooks and websockets, simple API key auth, organization-wide semantic search, structured data extraction, and usage-based pricing that scales with emails sent/received.

Here’s a demo of building an email agent: https://youtu.be/1V7BISeFUTM Here’s a demo of a voice agent with its own email inbox: https://youtu.be/eG2fCsRK4RY

So far AgentMail has been deployed to use cases such as apps with dedicated inboxes for each user, voice agents that receive documents in real time, automated account provisioning and QA testing, cold outbound platforms with thousands of inboxes, automations for processing invoices, and agents that coordinate work with humans and other agents.

We would love to hear your thoughts and feedback. You can try our playground at https://chat.agentmail.to

Your AI agent hits an infinite loop and racks up $2000 in API charges overnight. This happens weekly to AI developers.

AgentGuard monitors API calls in real-time and automatically kills your process when it hits your budget limit.

How it works:

Add 2 lines to any AI project:

  const agentGuard = require('agent-guard');
  await agentGuard.init({ limit: 50 }); // $50 budget

  // Your existing code runs unchanged
  const response = await openai.chat.completions.create({...});
  // AgentGuard tracks costs automatically

When your code hits $50 in API costs, AgentGuard stops execution and shows you exactly what happened.

Why I built this:

I got tired of seeing "I accidentally spent $500 on OpenAI" posts. Existing tools like tokencost help you measure costs after the fact, but nothing prevents runaway spending in real-time.

AgentGuard is essentially a circuit breaker for AI API costs. It's saved me from several costly bugs during development.

Limitations: Only works with OpenAI and Anthropic APIs currently. Cost calculations are estimates based on documented pricing.

Source: https://github.com/dipampaul17/AgentGuard

Install: npm i agent-guard

I've created and open-sourced an Astro developer blog template with an interactive theme changer that includes all 60 themes bundled with the JS code highlighter Shiki. Changing the theme affects the whole website including the code examples and Giscus comments. Inspired by the aesthetics of raw markdown, I wanted to create a beautiful blog like https://github.com/panr/hugo-theme-terminal but supercharged with a modern redesign and the incredible features of Astro.

Features:

  - Simple configuration file
  - Multiple theme modes (single, light/dark/auto, select)  
  - Giscus comments  
  - RSS feed  
  - SEO best practices + social card generation  
  - Markdown extensions (TOC, admonitions, reading time, etc)

I built a Chrome extension that acts as a WebSocket proxy, allowing real-time monitoring, message simulation, and traffic interception. Think "Proxyman for WebSockets" but integrated into Chrome DevTools.

Key features: Real-time WebSocket monitoring and message capture Send custom messages in both directions (client ↔ server) Block incoming/outgoing messages for testing Background monitoring (captures connections even when DevTools is closed) Why: I was debugging a WebSocket chat app and needed better tools than browser DevTools. Existing solutions required external proxies or were too basic.

Tech: Injects proxy script to intercept WebSocket constructor, React + Vite UI, Chrome DevTools API integration, MIT licensed. Perfect for debugging WebSocket apps, testing error scenarios, reverse engineering APIs, and QA testing real-time features.

Links:

GitHub: https://github.com/law-chain-hot/websocket-devtools

YouTube Demo: https://www.youtube.com/watch?v=L64x__1xORQ

Would love feedback from developers who work with WebSockets regularly!

Hi HN, Dylan and Shima here. Today we're launching the first version of our project, Safe Rate, an AI-native mortgage lender going all in on the chat experience, at saferate.com.

Mortgage shopping is an uncomfortable experience. You have to share lots of sensitive data to get a single quote, and if you want to comparison shop, the rate aggregator websites straight up sell your data, so you end up with a lot of texts, emails, and phone calls from salespeople you've never met asking you to blindly trust them with a big financial decision.

Getting a mortgage quote should be on your terms. You should be able to share your relevant details anonymously and get real pricing. You should know how competitive the offer is before speaking with someone. And you should still be able to connect with a human if you need support.

So we've built a suite of tools in our AI chat to put shoppers in control:

1. Instant mortgage quoting without sharing any sensitive data (currently live for conventional loans in Illinois)

2. Automatically review a Loan Estimate, which every lender must provide you by law, and see how the estimate scores (0 to 100, benchmarked against 2 million loans). We then can make an instant competing offer so you know if we can save you money.

3. For refinances, share a mortgage statement to see if you can save money today, and if not, create a personalized tracker so you can be notified when rates have fallen

4. Mortgage Q&A for commonly asked mortgage questions. Still lots more to add!

5. Live - a Zoom-like call for mortgage shopping that instantly connects you with a loan officer while our AI assistant listens in, updates the quote details, and gets you pricing

We support this functionality in a variety of ways:

1. Chat on https://saferate.com

2. Email at [email protected]

3. Live via our Zoom-like call tailored to mortgage shopping

4. Slackbot (please ask if you're interested in access)

5. Phone 1(888)850-6123

6. Coming soon RCS, SMS, and WhatsApp

Our chats are universal, so you can start with email and then transition to web chat, or vice versa, and it will retain the full context. And we've built our chat to be embeddable on any other website via an iframe. Safe Rate's main chat is actually an embedded iframe itself. And because we can't be wrong when giving a mortgage quote (regulation!) we've designed our AI so it does not hallucinate.

Here’s a 90 second demo video that shows this in action: https://youtu.be/6eIJukspmfU

You can also see a sample Live session: https://youtu.be/alAPtjUJ5xE

We're posting here today because any testing or feedback would be a great help. We're able to join the Live Call if you're brave enough (in addition to building this system ourselves, we're also licensed loan officers). And our quotes are only valid for Illinois with a max loan size of $806,500. We hope to expand nationally soon. Thanks!

Just small prototype showcasing diffusion models

Hey HN! I'm Mike, based in Stockholm/Sweden and I'm the founder of Zero Waste Cloud (https://zerowastecloud.io) - a tool that scans AWS and GCP infrastructure to find cost optimization opportunities while calculating the environmental impact. Almost all businesses can save 20-40% of their cloud costs using this tool (sources at bottom)

TLDR: Sign up -> Connect your AWS/GCP -> Scan -> Save 20-40% of your cloud spend.

My backstory: I've been a CISO and in IT management for a long time and always been frustrated by how much waste I'd find in every environment - idle EC2 instances running 24/7, oversized RDS databases, forgotten storage volumes, test resources that never gets removed. What also bothered me was that nobody was tracking the environmental cost of this waste. Every unused resource burns electricity and contributes to carbon emissions unnecessarily.

What it does:

- Scans AWS accounts using IAM roles or access keys, GCP projects using service account keys

- Identifies specific optimization opportunities: unused EC2/Compute Engine instances, oversized databases, unattached storage, missed reservation opportunities and more - Calculates financial savings AND CO₂ reduction for each recommendation using region-specific grid emission factors

- Multi-account/project support

- Generates detailed reports with prioritized recommendations

Technical details: Built with React/TypeScript frontend, Supabase backend. Uses AWS SDK for EC2, RDS, Cost Explorer APIs and GCP's Compute Engine, Cloud SQL, and Cloud Billing APIs. Carbon calculations combine cloud provider PUE data with regional electricity grid emission factors from government sources.

Try it out: The onboarding from creation to your first scan being run is only ~30 seconds. The scanning process is fully automated and typically completes in 5-30 minutes depending on account size of course.

I would genuinely love to get your feedback on it, just came out of beta a few days ago so if there's any bugs around please do let me know =)

If you're on Linkedin let's connect: https://www.linkedin.com/in/almstedt/

Sources:

"companies estimate that 21-50% of their cloud expenditure is wasted" (https://www.techmonitor.ai/hardware/cloud/cloud-waste-hits-b...)

"45% of cloud customer’s expenditures are spent on resources they will never use" (https://www.sciencedirect.com/science/article/abs/pii/S22105...)

"21% of enterprise cloud infrastructure spend /.. / in 2025*—is wasted on underutilized resources (https://www.prnewswire.com/news-releases/44-5-billion-in-inf...)

Hello HN!

I realized how often my mind wanders while waiting for Claude to finish "thinking". So I built Clauditate, a lightweight meditation menubar app that helps you with some breathing exercises.

It uses Claude Code hooks and pops up with a breathing guide. I added some heuristics to make it less "annoying", for example if you dismiss it twice within 30 mins, it won't show up for another 2 hours. You can also tweak this behavior (and more) in the settings.

It's early, but I'd love to hear your feedback!

Github: https://github.com/ykumards/clauditate

Hi HN, I've been using Claude to generate landing pages and it's a huge time-saver.

But, it's hard to go from Artifact -> Published website. That's where Artifact Ninja comes in.

How it works: 1. Click "Copy" above your artifact in Claude.ai and paste into Artifact.Ninja 2. We'll generate a static webpage without any Claude branding 3. Use your registrar's "masked forwarding" feature to connect a custom domain

Artifact Ninja is a quick utility I needed for my own purposes. It's cheap to host (Cloudflare worker + D1), so it's free! Hope it helps.

Hi HN!

I built and just launched DevNotify — an open-source platform to help developers keep track of their starred GitHub repositories and open issues, so they never miss important updates or contribution opportunities.

Why I built it: As a developer, I often star GitHub repos or open issues I care about, only to forget them later. GitHub doesn’t make it easy to track changes across multiple projects you're interested in. I wanted a way to centralize updates from projects I follow or contribute to — so I built DevNotify.

What DevNotify does: - Shows updates from your starred repositories - Tracks issues. - Great for finding contribution opportunities - Fully open-source and privacy-friendly

Live App: https://devnotify.in/app GitHub Repo: https://github.com/yourusername/devnotify

Would love to hear your thoughts and feedback — technical or UX-wise. I’m actively working on it, so open to suggestions, ideas, and contributions!

Thanks

Hey HN,

I’m a solo builder, and I made Brandkiit after getting stuck too many times trying to name and brand new ideas. I’d get excited about a side project, but before I could start building, I’d lose days (sometimes a whole week) trying to find:

• a name that felt right • an available domain • a decent logo • matching colors and fonts

…all before I wrote a single line of code.

Brandkiit is a tool that takes your idea and instantly gives you a full brand kit — including:

• name + domain availability • logo • color palette • font pairing • CSS file export you can send to your favourite code tool

It’s meant to help indie hackers, designers, and product folks get started without getting stuck in the branding spiral.

Would love your feedback! Happy to answer any questions.

Nima

I'm Craig Buckler and Publican is my tiny, simple, fast, and free static site generator for Node.js.

I've used several SSGs including Jekyll, Metalsmith, and Eleventy. Why build another? The main reason: personal preference. All SSGs have features that I need, features I don't need, and features they don't support. Publican implements just enough with flexibility to extend it using JavaScript.

Publican templates use JavaScript literal ${ expressions } so there's no weird syntax to learn. You can also use !{ expressions } to output partially-built pages for runtime use in Express.js or elsewhere. Features include:

- process any content: markdown, HTML, CSS, JavaScript, TXT, SVG, RSS, XML, etc.

- simple JavaScript configuration

- clean URL routing

- automated navigation, pagination, directory, and tag index pages

- built-in syntax highlighting

- virtual content and templates (passed as strings)

- extendable function hooks

- use whatever client-side framework you like (or none!)

- fast site build and file watch rebuild

- full documentation at https://publican.dev/docs/

- starter themes at https://publican.dev/themes/

You can install Publican using npm: https://www.npmjs.com/package/publican

The code is available at: https://github.com/craigbuckler/publican

Also available for Publican:

- https://publican.dev/livelocalhost/ - a hot-reloading development server

- https://publican.dev/staticsearch/ - a search engine for any static site

All feedback is appreciated!

Hey HN,

I'm Marcus, an open-source enthusiast, and I'm excited to share a new project!

I recently faced a common challenge: building a reporting page that needs to fetch and process data from a remote source daily, providing a simple UI for various KPIs.

While `prisma-queue` has served me well in other projects, its tie to the Prisma ORM wasn't ideal for this new use case, where the ORM choice was still open.

Many popular job queues are Redis-based, like BullMQ. However, I felt there was a gap for a robust, database-backed queue that leverages PostgreSQL's reliability without being locked into a specific ORM. And so, Vorsteh-Queue was born!

While there is currently no official "1.0" release yet, I believe Vorsteh Queue is ready for community feedback to help shape the future (code & docs) and accelerate the v1 launch.

Key features that might interest you:

* UTC-first timezone design: All timestamps are stored as UTC.

* Full TypeScript support with generic job payloads: Your job handlers are properly typed.

* ORM-agnostic with adapters for Drizzle and Prisma (PostgreSQL): More adapters are planned.

* Advanced scheduling: Includes priority queues, delayed jobs, cron scheduling, and progress tracking.

Vorsteh-Queue offers a compelling alternative for those who prefer to manage their background jobs directly within PostgreSQL, providing strong transactional guarantees and eliminating the need for a separate Redis instance.

Why the name "Vorsteh-Queue"?

The name is a tribute to my German Spaniel, a "Vorstehhund" or pointing dog. Just as a pointing dog indicates its target, Vorsteh-Queue aims to reliably point your application towards efficient background job processing. It's a nod to the personal passion that drives open-source development, inspired by stories like Bruno, the API client!

Try it out and let me know what you think! All feedback, contributions, and ideas are highly welcome.

Quick Links:

* Github repo: https://github.com/noxify/vorsteh-queue

* Docs: https://vorsteh-queue.dev/

Hello from the SWE-bench/SWE-agent team at Princeton/Stanford.

When we created the SWE-bench benchmark in 2023 from hundreds of real-life GitHub issues/pull requests, the highest score was just a couple of percent. The tasks were so challenging for LMs, that most people didn't even want to work on them.

Half a year later, SWE-agent showed that the early 2024 LMs were actually good enough to resolve up to 20% of the GitHub issues in the benchmark. This kicked off a whole wave of coding agents.

Back then, developing agents was all about working around tons of silly behavior from the LMs. For example, if a command didn't work, they would try running the exact command again. If a command didn't return output, they would assume it never ran. They also couldn't get whitespace right in their edits, would get stuck into repetitive attempts and much much more.

So agents got pretty complicated to work around all of that bad LM behavior.

But now it's 2025, and LM companies have invested a whole lot of money to make their LMs really good at being agents.

So we asked two questions:

1. What's the simplest agent we can write that still scores near SotA? 2. How do LMs compare when we evaluate them using this simple agent?

Turns out, the agent can be very simple indeed! mini-swe-agent (https://github.com/SWE-agent/mini-swe-agent) has only 100 lines of code for the agent class (plus some 100 lines for environment etc.). It is little more than a loop that parses LM output for shell commands, executing them in a subshell, and continuing.

We then took various LMs and put them to the test in a real apples-to-apples comparison without a fancy agent scaffold to prop up bad LMs.

Our new leaderboard https://www.swebench.com/ shows the results.

The highest score is currently 65% with Claude Sonnet 4 (which is not much less than the 70% that most fancier agents observe).

o3, o4-mini, and Gemini 2.5 Pro are significantly behind, but not hopeless, achieving 50-60%.

We were really surprised by these strong numbers overall: It shows that as LMs get stronger and better adapted at performing difficult, highly iterative tasks, we can take our hands off of the steering wheel, provide the minimal necessary environment, and let the LM figure out the rest.

Let us know if you have any questions, our team is here on HN today :)

I open-sourced this tool to generate colorful waveform images from audio files, inspired by Pioneer/Denon gear. No FFT required, just filters and peak windows.

Hey HN,

  We built an Airbnb data API because we were tired of the unreliable and expensive options out there.

  Our goal was to make it fast, cheap, and easy for developers to get the data they need.

  We're seeing people use it to build property management tools, fuel investment models, create rental calculators, and analyze tourism trends.

  Check it out:
   - general: https://www.airroi.com/api
   - See docs: https://www.airroi.com/api/documentation

I've been an OmniFocus user for years but always felt something was missing – I could capture everything but couldn't see the bigger picture of my work. Traditional task managers are great at collection and to do lists but I think they don't solve the problem of visualising and managing work.

*What It Does*

Flowcus is a macOS app that visualises your OmniFocus, Apple Reminders & plain text tasks as a customisable Kanban board. It's not intended to replace your task managers, rather it enhances them by providing:

- Visual workflow with customisable columns and swimlanes - Drag-and-drop task management that syncs back to your original task managers - Work-in-progress limits to prevent overcommitment - Eisenhower matrix view for prioritising important vs. urgent work - Ability to mark tasks as blocked and visualise dependencies - A unified view of tasks from multiple platforms - Throughput & ageing work metrics

For example, you might have research tasks in OmniFocus, meeting reminders in Apple Reminders and basic tasks in plain text. Flowcus shows them all on one board, letting you drag tasks from "In Progress" to "Done" and having that status change sync back to the original app.

## Why It's Useful

Traditional task managers aren't great at showing you a holistic view of your work. Flowcus bridges this gap by applying lean thinking principles to personal productivity:

1. *Makes work visible* - See all your tasks in one place across different stages 2. *Limits work in progress* - Stop starting, start finishing with WIP limits 3. *Manages flow* - Identify and eliminate bottlenecks in your workflow 4. *Provides feedback* - See how work moves through your system over time

For OmniFocus power users, Flowcus adds the visualisation layer that's always been missing. For people juggling multiple task systems, it unifies your tasks without forcing you to use multiple apps. That's the idea anyway...

## How I Built It

I started with Python scripts to extract OmniFocus data, but realised I needed proper two-way sync and a native Mac experience. I used Claude Code to help me build the first version to integrate with OmniFocus's AppleScript interface and Apple Reminders' EventKit framework. The app monitors changes in real-time and propagates updates back to the source systems. The first prototype was rough but I could see the benefits straight away: I could visualise my work, arrange it into swimlanes, identify when I was overcommitting and track the flow of work across different stages.

## Current Status

The app is currently in beta (v0.2.5). I use it every day to help me manage my work. I can't do without it now. I'm considering adding more task manager integrations at some point. I'd love feedback from any fellow productivity geeks and OmniFocus users – thanks!

Hi HN! Going to meetups or conferences?

TL;DR: We’re building an AI tool that helps you capture leads, keep track of key details, and follow up fast starting with in-person events.

We’re Tonje and German, and we’re building Moots AI, a smart contact system for sales teams, founders, investors, recruiters, and partnership managers who rely on real-world events to meet leads, close deals, and build relationships.

Our first focus: make sure no contact or lead from a conference, summit, or meetup slips through the cracks.

The Problem: Most networking events are a waste.

You meet dozens of valuable people but the context fades fast, and follow-ups are often delayed or forgotten. If you’re in sales, fundraising, or biz dev, missed follow-ups = lost pipeline.

- 80% of leads generated at trade shows are never followed up. - 35-50% of sales go to the vendor that responds first after an event. - Companies that follow up within 48 hours see a 25–50% higher conversion rate. - After 2–3 days, interest drops sharply especially for leads from in-person events.

You’ve probably lived this:

You show up early, set up your booth, and wait while everyone’s in talks. Then suddenly, right after a session ends, the crowd rushes in. You're juggling five conversations at once. No time to scan badges. No time to take notes. After the event, it's a blur.

You vaguely remember someone from Minneapolis who mentioned tennis and his kid… but who was it? Your lead capture tool didn’t save their name, photo, or context. That opportunity? Gone. Customer? Partner? Hire? You don’t remember and you can’t follow up on what you’ve forgotten.

Ever bump iPhones with someone at an event then forget who they were the next day?

When someone shares their number by bumping iPhones, you’re left with a contact buried in your phonebook—no notes, no context, just a name and a number.

Our Solution: AI that captures and activates your relationships

Moots turns live conversations into rich, actionable contact profiles. Scan a badge or card → Moots instantly builds a profile with job title, company, LinkedIn, education

It transcribes the conversation, tags key details, and qualifies the lead It helps you prioritize follow-ups, suggests next steps, and can draft outreach for you It even surfaces on the fly, contacts in common, shared interests, or past locations to help you connect with context

Every person you meet becomes part of a searchable, AI-powered personal network. Your AI agent remembers everything so you can focus on building relationships and closing deals.

Demo: https://youtu.be/tkC1DCQzHcY?si=Vlu-5TEUNXWyE09K

What Moots does today

- Scan lead badges or business cards at events - Auto-enrich contacts with public info (title, LinkedIn, company, education) - Dictate and attach notes to the right contact - Qualify and categorize leads - Search your network by event, tag, or date - Draft follow-ups - Build a private, living knowledge base for each contact

Our Team

Tonjé Bakang Tonje (CEO) – Previously founded Afrostream (YC W15), a media platform that reached millions. Former VC and dealmaker in Europe.

German Saprykin (CTO) – Engineering leader with deep experience in AI infrastructure and consumer-scale systems. Previously at Grab.

We just launched, download Moots AI on the App Store:

https://apps.apple.com/us/app/moots-ai/id6717572977

We’re always learning. Feedback is gold. Tell us how we can better support your business and your relationships.

Tonje & German